αT-Catenin Is a Constitutive Actin-binding α-Catenin That Directly Couples the Cadherin·Catenin Complex to Actin Filaments*

α-Catenin is the primary link between the cadherin·catenin complex and the actin cytoskeleton. Mammalian αE-catenin is allosterically regulated: the monomer binds the β-catenin·cadherin complex, whereas the homodimer does not bind β-catenin but interacts with F-actin. As part of the cadherin·catenin complex, αE-catenin requires force to bind F-actin strongly. It is not known whether these properties are conserved across the mammalian α-catenin family. Here we show that αT (testes)-catenin, a protein unique to amniotes that is expressed predominantly in the heart, is a constitutive actin-binding α-catenin. We demonstrate that αT-catenin is primarily a monomer in solution and that αT-catenin monomer binds F-actin in cosedimentation assays as strongly as αE-catenin homodimer. The β-catenin·αT-catenin heterocomplex also binds F-actin with high affinity unlike the β-catenin·αE-catenin complex, indicating that αT-catenin can directly link the cadherin·catenin complex to the actin cytoskeleton. Finally, we show that a mutation in αT-catenin linked to arrhythmogenic right ventricular cardiomyopathy, V94D, promotes homodimerization, blocks β-catenin binding, and in cardiomyocytes disrupts localization at cell-cell contacts. Together, our data demonstrate that αT-catenin is a constitutively active actin-binding protein that can physically couple the cadherin·catenin complex to F-actin in the absence of tension. We speculate that these properties are optimized to meet the demands of cardiomyocyte adhesion.

The adherens junction (AJ) 2 mechanically couples the actin cytoskeletons of adjacent cells to establish and maintain intercellular adhesion (1)(2)(3). The core of the AJ is the cadherin⅐catenin complex (4). Classical cadherins are single pass transmembrane proteins with an extracellular domain that mediates calcium-dependent homotypic interactions (5). The adhesive properties of classical cadherins are driven by the recruitment of cytosolic catenin proteins to the cadherin tail: p120-catenin binds to the juxtamembrane domain, and ␤-catenin binds to the distal part of the tail (6). ␤-Catenin, in turn, recruits ␣-catenin to the cadherin⅐catenin complex (7,8). ␣-Catenin is a filamentous actin (F-actin)-binding protein and the primary link between the AJ and the actin cytoskeleton (9 -12).
In mammals, ␣E (epithelial)-catenin is allosterically regulated: the monomer binds the ␤-catenin⅐cadherin complex, whereas the homodimer does not bind ␤-catenin but interacts with F-actin (9,10). ␤-Catenin binding to ␣E-catenin sterically hinders F-actin binding (8,13), explaining how ␣E-catenin as part of the cadherin⅐catenin complex has a weak affinity for F-actin. More recently, it was shown that the cadherin⅐catenin complex binds strongly to F-actin under force, indicating that the ␣E-catenin-actin interface is dynamically regulated by tension (12). In addition, evidence suggests that tension can regulate ␣E-catenin conformation: actomyosin-generated force stretches the middle (M) domain to reveal binding sites for cytoskeletal proteins such as vinculin (14 -18). Thus, ␣Ecatenin is a dynamic and multifunctional protein regulated by tension.
␣-Catenin functions in adhesion and mechanical signaling must be integrated in all tissues. In cardiomyocytes, the AJ functions with the desmosome to physically link opposing cells in a specialized adhesive structure called the intercalated disc (ICD) (19). Contractile forces place physical demands on heart junctional complexes: not only must they withstand repeated cycles of force, but tension-sensing proteins within these complexes must be "tuned" to regulate signaling and maintain homeostasis (20). Two ␣-catenin proteins are expressed in the mammalian heart, ␣E-catenin and ␣T (testes)-catenin (21)(22)(23). In contrast to the widely studied and well defined mammalian ␣E-catenin, little is known about ␣T-catenin, a protein unique to amniotes that is expressed predominantly in the heart and testes (22,24). ␣T-Catenin is expressed in cardiomyocytes where it localizes to the ICD, and genetic ablation of ␣T-catenin in mice causes dilated cardiomyopathy (22,23,25). Notably, two mutations in ␣T-catenin have been linked to arrhythmogenic right ventricular cardiomyopathy (ARVC): an amino acid (aa) change in the N terminus (valine to aspartic acid, V94D) and deletion of one aa in the C-terminal ABD (loss of a leucine, L765del) (26). However, the molecular properties of ␣T-catenin are undefined, and how these mutations affect ␣T-catenin function in cardiomyocytes remains unclear.
Here we show that ␣T-catenin is a constitutive actin-binding ␣-catenin that can directly couple the AJ to the actin cytoskeleton. Our data also reveal that the V94D mutation linked to ARVC alters ␣T-catenin dimerization potential to disrupt ␤-catenin binding and cellular localization. We postulate that ␣T-catenin protein conformation and ligand binding proper-ties are tuned to meet the specific demands of cardiomyocyte adhesion.

␣T-Catenin Is a Constitutive Actin-binding Protein
binds ␤-catenin, and mediates homodimerization (7). The M region is composed of three four-helix bundles (Fig. 1A, M1-M3) and binds vinculin in response to mechanical force (14 -17, 29, 30). A small linker region connects the C-terminal five-helix actin-binding domain (ABD) to the M region (Fig.  1A). We compared the amino acid sequence of Mus musculus ␣T-catenin with M. musculus ␣E-catenin and M. musculus ␣N-catenin. ␣T-Catenin is 58% identical and 77% similar to ␣E-catenin; likewise, it is 59% identical and 77% similar to ␣N-catenin (Fig. 1B). ␣E-Catenin and ␣N-catenin are 81% identical and 91% similar, making ␣T-catenin the most divergent of the mammalian family. We then analyzed sequence homology across domains between ␣T-catenin and ␣E-catenin (Fig. 1C). The region with the lowest degree of homology is N2 (39% identical and 61% similar), whereas the region with the highest degree of homology is M2 (62% identical and 92% similar).
We then questioned whether sequence differences affected domain organization in ␣T-catenin. We purified recombinant M. musculus ␣T-catenin and M. musculus ␣E-catenin from Escherichia coli and used limited trypsin proteolysis to examine domain organization. As shown previously (31,32), tryptic digestion of ␣E-catenin monomer revealed two stable fragments: the modulation domain (aa 385-651) and the ␤-catenin-binding/homodimerization domain (aa 82-287) (Fig. 1D). Tryptic digestion of ␣T-catenin revealed three stable fragments at 30, 25, and 18 kDa (Fig. 1D). N-terminal sequencing revealed that the 30-kDa fragment started at aa 379 and contained bundles M2 and M3 (Fig. 1D). The entire M2-M3 region forms a protease-resistant fragment in mouse ␣E-catenin ( Fig. 1D) (10,31,33) and fish ␣E-catenin (32). Notably, the 18-kDa fragment started at aa 485, near the end of domain M2, and contained the entire M3 domain. This suggests that, unlike ␣E-catenin, the ␣T-catenin M2-M3 region exists in a more open, protease-sensitive state. Finally, the 25-kDa fragment started at aa 108, similar to the dimerization/ ␤-catenin-binding domain in ␣E-catenin (aa 82-287), although this fragment, similar to M2-M3, was markedly less proteaseresistant than in ␣E-catenin. We conclude that the conformation of ␣T-catenin is similar to ␣E-catenin but with differences in the stability of both N-terminal and middle domains that could impact function.
␣T-Catenin Is a Monomer in Solution-We assessed the oligomerization state of ␣T-catenin by chromatography. Recombinant ␣T-catenin protein prepared from E. coli was first purified by Mono Q ion exchange chromatography ( Fig. 2A). Two peaks were routinely observed during elution off a Mono Q column ( Fig. 2A, top chromatogram), and SDS-PAGE analysis of peak fractions revealed they both contained full-length ␣T-catenin ( Fig. 2A, bottom gel). A similar ion exchange chromatography profile is observed with M. musculus ␣E-catenin (data not shown), and the two peaks correspond to the monomer (peak 1) and homodimer (peak 2) species. Both ␣T-catenin peak fractions were subsequently purified over a Superdex 200 (S200) size exclusion chromatography (SEC) column. The Mono Q peak 1 fraction eluted in a single, discrete peak (Fig. 2B, purple line), consistent with it being a single, likely monomeric, species. The S200 elution profile of Mono Q peak 2 was similar to peak 1, although a second, small peak was sometimes observed where a dimer species would be expected to elute (Fig. 2B, red line).
We then compared the primary S200 peak (elution volume, 60 -70 ml; concentrated to 25-50 M) of ␣T-catenin with ␣E-catenin monomer and homodimer by analytical SEC. At all concentrations tested (25-50 M), ␣T-catenin eluted in a single peak after both ␣E-catenin homodimer and monomer, suggesting that ␣T-catenin is a monomer (Fig. 2C). We then used SEC and sucrose density gradient centrifugation to determine the molecular mass of ␣T-catenin, ␣E-catenin monomer, and ␣E-catenin homodimer (34). The SEC elution profiles (Fig. 2C) were compared with known standard proteins to calculate the Stokes radius (Fig. 2D). The calculated Stokes radius of ␣E-catenin homodimer was similar to past observations (6.5 versus 7.4 nm; Ref. 35), and the Stokes radii of both ␣E-catenin monomer and homodimer species were comparable with our previously measured radii of gyration from small angle x-ray scattering (4.5 and 6.0 nm, respectively; Ref. 32). The Stokes radius of ␣T-catenin was calculated to be 4.7 nm, slightly smaller than that of ␣E-catenin monomer (Fig. 2D).
We then used sucrose density gradient centrifugation to determine the sedimentation coefficients of ␣T-catenin, ␣E-catenin monomer, and ␣E-catenin homodimer. Proteins were separated on 5-20% sucrose gradients, and the fraction peak was determined and compared with a standard curve to calculate the sedimentation coefficient (Fig. 2, E and F). The Svedberg coefficients were determined to be 7.0S for ␣E-catenin homodimer (identical to past calculation (35)), 5.2S for ␣E-catenin monomer, and 5.7S for ␣T-catenin. Molecular masses were then estimated based on the measured Stokes radii and sedimentation coefficients (Fig. 2G). The molecular mass of ␣T-catenin was calculated to be 109 kDa, similar to that of ␣E-catenin monomer (106 kDa). Finally, ␣T-catenin migrated as a single band by native PAGE, faster than either ␣E-catenin monomer or dimer, consistent with the SEC analysis (Fig. 2H). We conclude that ␣T-catenin is primarily a monomer in solution.
Dimerization kinetics differ significantly between mouse ␣E-catenin and ␣N-catenin at physiological temperatures (8). ␣E-Catenin homodimerization is significantly weaker than ␣N-catenin homodimerization, but a kinetic block limits disassociation once an ␣E-catenin dimer is formed. The presence of two peaks in the Mono Q elution profile ( Fig. 2A) and the minor peak in the peak 2 SEC elution (Fig. 2B) suggest that ␣T-catenin might exist as a homodimer. However, if the Mono Q peak 2 elution represented a homodimer species of ␣T-catenin, then the majority of these dimers dissociated during SEC (Fig. 2B). We were never able to purify a sufficient quantity of the potential dimer species for analysis by SEC or native-PAGE. Also, attempts to promote dimerization by incubation of the monomer at physiological (37°C) temperatures caused the protein to aggregate and fall out of solution. Although we were unable to analyze the dimerization kinetics of wild-type (WT) ␣Tcatenin, our analysis of the V94D mutant revealed that ␣T-catenin, similar to ␣E-catenin and ␣N-catenin, has dimerization potential (described below). Nonetheless, we took advantage of the lack of a stable dimer in solution to study the behavior of ␣T-catenin monomer binding to F-actin.
␣T-Catenin Monomer Binds F-actin-Mammalian ␣Ecatenin binds and bundles F-actin (9 -12, 36), although in the absence of force, homodimerization is required to potentiate F-actin binding. We tested whether ␣T-catenin monomer binds F-actin using an F-actin cosedimentation assay. Increasing concentrations of ␣T-catenin were incubated in the pres-ence or absence of 2 M F-actin, the samples were centrifuged, and the resulting pellets were analyzed. ␣T-Catenin cosedimented with F-actin above background (Fig. 3A), and the bound protein was quantified and plotted over free protein to calculate the affinity of the interaction (Fig. 3B). Bovine serum albumin (BSA) and ␣E-catenin were run as negative and positive controls, respectively (Fig. 3A, right panels). Plotted data were fit to a hyperbolic function (Fig. 3B). ␣T-Catenin bound to  36). Thus, ␣T-catenin monomer is a constitutive actin-binding protein, and unlike ␣E-catenin, homodimerization is not required for strong F-actin binding in the absence of force (9,10,12).

␣T-Catenin Is a Constitutive Actin-binding Protein
To investigate whether ␣T-catenin monomer bundles F-actin, we used transmission electron microscopy to visualize ␣T-catenin incubated with actin filaments. Weak bundling of 2 M F-actin was observed with 4 M ␣T-catenin ( Fig. 3C and quantification in Fig. 5D). In contrast, robust bundling of 2 M F-actin was observed with 4 M ␣E-catenin homodimer (Figs. 3C and 5D). The weak bundling observed with ␣T-catenin could result from either the dimer species being stabilized on the actin filament or activation of a cryptic dimerization domain as observed in the vinculin tail (37). We conclude that ␣T-catenin is a poor bundler of F-actin.
␣T-Catenin Couples ␤-Catenin to F-actin-Binding to ␤-catenin weakens the affinity of ␣E-catenin for F-actin (9, 10). To test whether ␣T-catenin can bind F-actin as part of the cadherin⅐catenin complex, we purified mouse ␤-catenin and mixed it with ␣T-catenin. As expected, ␣T-catenin bound to ␤-catenin with a 1:1 stoichiometry (data not shown), and we isolated the ␤-catenin⅐␣T-catenin complex by SEC. Increasing concentrations of the ␤-catenin⅐␣T-catenin complex were incubated in the presence or absence of F-actin and centrifuged, and the pelleted material was analyzed as above. Although the ␤-catenin⅐␣T-catenin complex pelleted in the absence of F-actin (Fig. 3D, No F-actin panel), we were able to calculate the affinity of the complex for F-actin. The ␤-catenin⅐␣T-catenin complex bound to F-actin with a K d of 1.1 Ϯ 0.2 M (Fig. 3E). Although ␤-catenin lowers the affinity of ␣T-catenin for F-actin slightly, the interaction strength is considerably stronger than that of the Danio rerio ␤-catenin⅐␣Ecatenin complex (Ͼ10 M) and similar to the strength of ␣E-catenin homodimer association with F-actin (32,36). Thus, ␣T-catenin can bind both ␤-catenin and F-actin simultaneously to directly link the cadherin⅐catenin complex to the actin cytoskeleton. This is distinct from ␣E-catenin in which force is needed to strengthen the association between the cadherin⅐catenin complex and F-actin (12). Although tension may strengthen the interaction between ␣T-catenin and F-actin, we speculate that basal binding permits coupling between the cadherin⅐catenin complex and actin through ␣T-catenin over a range of forces.
␣T-Catenin V94D Mutation Creates an Obligate Homodimer-Two mutations in ␣T-catenin have been linked to ARVC: replacement of a valine for an aspartic acid at aa 94 (V94D) in the N1 domain and deletion of a leucine at aa 765 (L765del) in the ABD (26). Yeast two-hybrid and overexpression studies suggest that the V94D mutant interferes with ␤-catenin binding and that the L765del mutation promotes oligomerization (26). However, it is not clear how these mutations affect the biochemical properties of ␣T-catenin or impact cellular function in cardiomyocytes. We used site-directed mutagenesis to make the V94D and L765del mutations in ␣T-catenin and attempted to purify the mutant proteins. We were unable to purify L765del; the mutation rendered the expressed protein insoluble (data not shown). However, we were successful in expressing and purifying the V94D mutant. Surprisingly, V94D eluted as a single peak off the Mono Q column rather than two as observed with WT ␣T-catenin (Fig.  4A). We then ran the V94D peak over an S200 SEC column where it eluted as a single peak before WT ␣T-catenin and similar to the possible homodimer peak (Fig. 4B). We then compared the concentrated V94D protein (25-30 M; concentrations greater than this precipitated out of solution) with WT ␣T-catenin by analytical SEC. The V94D mutant eluted as a single species before WT ␣T-catenin with a larger Stokes radius (Fig. 4, C and G; 5.8 versus 4.7 nm). The V94D mutant also displayed a higher sedimentation coefficient than WT ␣T-catenin (Fig. 4, D and G; 7.7S versus 5.7S). The Stokes radius and sedimentation coefficient produced a molecular mass of 183 kDa (Fig. 4G), roughly double that of WT ␣T-catenin. We conclude that the V94D mutation creates a stable ␣T-catenin homodimer.
Because full-length ␣T-catenin V94D is difficult to purify, we deleted the ABD (aa 660 -895) in both WT and V94D ␣T-catenin to improve protein yield. We analyzed the SEC and sedimentation properties of the ⌬ABD constructs (Fig.  4, E-G). Similar to the full-length construct, the V94D mutation altered the elution and sedimentation profiles of the ⌬ABD construct (Fig. 4, E and F). The calculated molecular mass of ␣T-catenin V94D ⌬ABD was 146 kDa compared with 90 kDa for ␣T-catenin ⌬ABD, consistent with it forming a homodimer.
We analyzed the oligomeric state of the ␣T-catenin ⌬ABD proteins by cross-linking. Increasing concentrations of ␣Tcatenin ⌬ABD and ␣T-catenin V94D ⌬ABD were incubated with or without the cross-linker bis(sulfosuccinimidyl)suberate (BS3), and the resulting products were analyzed by SDS-PAGE. As expected, ␣T-catenin ⌬ABD and ␣T-catenin V94D ⌬ABD ran as 75-kDa proteins in the absence of cross-linker (Fig. 4H).
In the presence of BS3, however, V94D migrated as a 150-kDa protein at all concentrations tested, indicating a cross-linked  JULY 22, 2016 • VOLUME 291 • NUMBER 30 dimer. Incubation with BS3 did not affect ␣T-catenin ⌬ABD migration at low concentrations, although at higher concentrations (2 and 4 M), a 150-kDa species was detected. We speculate that this could reflect a transient homodimer species. We conclude that the V94D mutation promotes dimerization of ␣T-catenin.

␣T-Catenin Is a Constitutive Actin-binding Protein
We used limited proteolysis to determine whether the V94D mutation affected domain organization. Like WT ␣T-catenin,

␣T-Catenin Is a Constitutive Actin-binding Protein
three fragments were resistant to trypsin cleavage in V94D (Fig.  4I). However, the ␤-catenin/homodimerization domain (aa 108 start; confirmed by Edman degradation sequencing) was protected relative to WT (Fig. 4I, blue arrowhead, compare with Fig. 1D, blue arrowhead), consistent with this domain being stabilized in the homodimer state. We then questioned whether the V94D homodimer could interact with ␤-catenin. We mixed increasing concentrations of WT or V94D ␣T-catenin with GST-tagged ␤-catenin, pulled down the ␤-catenin, and assessed binding. Wild-type ␣T-catenin bound GST-␤-catenin at stoichiometric levels; however, little to no V94D bound (Fig. 4J). Thus, the V94D mutation creates an obligate ␣T-catenin homodimer that cannot bind ␤-catenin.
We then tested whether the V94D homodimer could bundle F-actin. We consistently observed increased bundling of 2 M F-actin with 4 M ␣T-catenin V94D relative to 4 M WT ␣T-catenin (Fig. 5, C and D). Although increased, the level of bundling was still less than that observed with 4 M ␣Ecatenin homodimer (Figs. 3C and 5D). We conclude that the V94D mutation promotes ␣T-catenin-mediated F-actin bundling.
␣T-Catenin V94D Disrupts Localization in Cardiomyocytes-␣T-Catenin localizes to the adherens junction at the ICD in cardiomyocytes (22). To determine whether the V94D mutation disrupted ␣T-catenin cellular localization, we transiently expressed EGFP-tagged WT or V94D ␣T-catenin in neonatal mouse cardiomyocytes. EGFP-␣T-catenin localized specifically to cell-cell contacts in cardiomyocytes where it colocalized with both ␣E-catenin and N-cadherin (Fig. 6, A, C, and zoom in E). In contrast, V94D was largely peripheral to cell-cell contacts (Fig. 6, B, D, and zoom in E) and localized to actin fibers (Fig. 6, B and D, orange arrowheads). This was confirmed by directly measuring colocalization between N-cadherin and EGFP-␣T-catenin or EGFP-␣T-catenin V94D signals at AJ clusters in transfected cells using Pearson's r (Fig. 6F). This analysis revealed a significant reduction in colocalization between N-cadherin and EGFP-␣T-catenin V94D at AJs (Fig.  6F). Thus, the V94D mutation disrupts ␣T-catenin subcellular localization in cardiomyocytes.

Discussion
␣T-Catenin Binds F-actin Strongly as a Monomer-Our in vitro results show that, in solution, ␣T-catenin binds F-actin as a monomer and in complex with ␤-catenin, properties that separate it from mammalian ␣E-catenin. ␣T-Catenin monomer binds F-actin with a slightly higher affinity than ␣E-catenin homodimer (0.4 versus 1.0 M) (36). Although ␤-catenin binding reduces the affinity of ␣T-catenin for F-actin, the reduction is relatively small (from 0.4 to 1.1 M). We conclude that ␣T-catenin binding to F-actin, unlike mammalian ␣E-catenin, is not allosterically regulated. This would permit ␣T-catenin to directly couple the cadherin⅐catenin complex to the actin cytoskeleton in the absence of tension, although mechanical force could strengthen the ␣T-catenin-actin interface.
␣T-Catenin Has Dimerization Potential-Both M. musculus ␣E-catenin and ␣N-catenin homodimerize in solution, although the kinetics of dimerization differ significantly between the two mammalian ␣-catenins (8). At physiological temperature, the homodimerization affinity of ␣N-catenin is more than 10ϫ greater than the homodimerization affinity of ␣E-catenin (2 versus 25 M). However, the kinetics of dissociation differ markedly: ␣N-catenin equilibrates quickly, whereas a kinetic block limits ␣E-catenin dissociation (8). The ␣E-catenin dimer is thus stabilized and can persist at concentrations well below the K d of association. Our in vitro results suggest that ␣T-catenin has the ability to homodimerize. We observed a monomer and putative dimer species by ion exchange chromatography, although the dimer quickly dissociated upon dilution during SEC. Stronger evidence comes from our analysis of the V94D mutation where a single amino acid change shifted the protein to the homodimer state. Cross-linking studies with the ␣T-catenin ⌬ABD constructs also provide evidence for dimerization potential in the WT protein. Unfortunately, our inability to maintain soluble ␣T-catenin at or near physiological temperature (37°C) precluded a detailed analysis of dimerization kinetics. Nonetheless, our results lead us to postulate that ␣T-catenin has dimerization potential and that the homodimer species, similar to ␣N-catenin, dissociates quickly (i.e. no kinetic block).
Evidence suggests a potential role for the ␣-catenin homodimer in migration and cell-cell adhesion (36,38,39). However, a physiological role for the ␣-catenin homodimer in cardiomyocytes and whether putative ␣E-catenin and ␣Tcatenin homodimers function similarly in vivo are unclear. The V94D mutation, which drives ␣T-catenin into the dimer state in vitro, shifted localization from cell-cell contacts and pro-

␣T-Catenin Is a Constitutive Actin-binding Protein
moted recruitment to F-actin bundles when expressed in cardiomyocytes. Actin filament cross-linking is essential for cardiomyocyte cytoskeletal organization and function. The barbed ends of actin filaments from adjoining sarcomeres interdigitate at the Z-disc where they are cross-linked primarily by ␣-actinin to form a structural lattice (40). ␣-Actinin is an established ␣-catenin ligand (41,42), and we have detected ␣T-catenin in complex with ␣-actinin in cardiomyocyte lysates. 3 Thus, the ␣T-catenin homodimer could have a role in cytoskeleton organization in cardiomyocytes. Alternatively, homodimerization may serve to regulate interactions with ␤-catenin and/or plakoglobin along the ICD. Additional work is needed to elucidate the putative role of the ␣T-catenin homodimer in cardiomyocyte biology.
V94D Mutation Linked to ARVC Promotes Homodimerization-The V94D mutation in ␣T-catenin is linked to ARVC, although the heterozygous mutation has only been documented in one individual (26). It was shown previously that the mutation reduced both ␤-catenin binding and homodimerization potential in a yeast two-hybrid assay (26). In contrast, we found that V94D promotes ␣T-catenin homodimerization, in effect creating an obligate homodimer species that cannot bind ␤-catenin. Not surprisingly, the V94D mutant disrupted cell-cell contact localization when expressed in cardiomyocytes. In the heterozygous state, it is unclear whether 1) V94D interacts with WT ␣T-catenin to disrupt localization to cell junctions and ␣T-catenin-mediated adhesion and/or 2) the mislocalized mutant protein disrupts cytoskeletal organization. Nonetheless, to the best of our knowledge, this is one of the first demonstrations of how a disease-linked mutation in ␣-catenin disrupts a fundamental molecular property.
␣T-Catenin Domain Stability-Our limited proteolysis experiments revealed that both the ␤-catenin/homodimerization domain and middle domain were more protease-sensitive in ␣T-catenin than in ␣E-catenin. Notably, the N2 bundle within the ␤-catenin/homodimerization domain of ␣T-catenin is the region with the least conservation compared with ␣E-catenin. ␣T-Catenin binds ␤-catenin (Fig. 4D) and plakoglobin, 4 although the strengths of these interactions are untested. Differences in N2 could impact ␣T-catenin ligand binding, including self-association, to regulate molecular complex formation at cell-cell contacts.
The core M region (M1-M3) of ␣E-catenin is required for its function as a mechanosensor in which tension alters ␣-catenin conformation to promote ligand binding (14,16,29,43). Recent structural and single molecule studies coupled with molecular dynamics simulations support a model in which mechanical force reorients M2 and M3 to release M1, which contains the    Fig. 2A). B, S200 SEC of ␣T-catenin V94D Mono Q peak fraction and ␣T-catenin WT peak 2 fraction. C, analytical S200 SEC of ␣T-catenin V94D and ␣T-catenin WT. The elution profile was used to calculate R S in G. D, sucrose gradient sedimentation of ␣T-catenin V94D. Fractions were collected from 5-20% sucrose gradients and analyzed by Coomassie-stained SDS-PAGE. The percentage of V94D in each fraction was measured and plotted, and the data were fit to a Gaussian curve (red line). The ␣T-catenin sedimentation profile from Fig. 2E (dashed orange line) is shown for comparison. The fraction peak was used to calculate the sedimentation coefficient in G. E, analytical S200 SEC of ␣T-catenin (␣T-cat) ⌬ABD and ␣T-catenin V94D ⌬ABD.

␣T-Catenin Is a Constitutive Actin-binding Protein
vinculin-binding domain (16,17,29). A salt bridge network between M domains is postulated to maintain ␣E-catenin in the autoinhibited conformation in the absence of tension (17). Based on sequence homology, a similar salt bridge network could exist in ␣T-catenin, although our limited proteolysis results showed that the ␣T-catenin M fragment (M2-M3) was less stable than in ␣E-catenin. We speculate that increased flexibility within the ␣T-catenin M2 and M3 domains could reduce the force required for activation, permitting M1 release and ligand recruitment at lower tension states. Increased flexibility between the M2 and M3 domains could also promote ligand binding within this region. Notably,

␣T-Catenin Is a Constitutive Actin-binding Protein
␣T-catenin, but not ␣E-catenin, was shown to bind plakophilin-2, a desmosomal protein that links to intermediate filaments, and the binding interface was mapped to M3 (23). ␣T-Catenin, through association with plakophilin-2, may function as a molecular link to integrate the actin and intermediate filament cytoskeletons at the ICD. It is possible that structural differences within the core M region between ␣-catenins could regulate both mechanosensing and ligand binding properties.
␣T-Catenin Function in Cardiomyocytes-␣-Catenin functions in adhesion and mechanical signaling must be integrated in all tissues. Contractile forces place physical demands on heart junctional complexes: not only must they withstand repeated cycles of force but tension-sensing proteins within these complexes must be tuned to regulate signaling and maintain homeostasis. Our in vitro studies showed that ␣T-catenin could directly couple the actin cytoskeleton to cadherin⅐catenin

␣T-Catenin Is a Constitutive Actin-binding Protein
in the absence of tension. We speculate that this property of ␣T-catenin might permit the cadherin⅐catenin complex to maintain a static linkage to the actomyosin network over a range of forces such as those produced by repeated cycles of contraction and relaxation in cardiomyocytes. Our biochemical analyses also suggest that ␣T-catenin dimerization properties and M region stability differ from those in ␣E-catenin. How these differences impact in vivo function is unclear, but we speculate that they could impact molecular interactions and tension sensing. In the mammalian heart, ␣T-catenin may have evolved to complement ␣E-catenin functions in adhesion and signaling.

Experimental Procedures
Plasmids-DNA encoding full-length M. musculus ␣Tcatenin was cloned into pGEX-TEV (36) to create a fusion between GST and ␣T-catenin. Site-directed mutagenesis was used to create the valine to aspartic acid mutation at amino acid 94 (V94D) in ␣T-catenin. The N-terminal head region (aa 1-659) of ␣T-catenin or ␣T-catenin V94D was cloned into pGEX-TEV to create the ⌬ABD constructs. WT and V94D ␣T-catenin were cloned into pEGFP-C1 for expression in mammalian cells.
Recombinant Protein Expression and Purification-GSTtagged ␣T-catenin, ␣E-catenin, and ␤-catenin were expressed in BL21(DE3) E. coli cells and purified as described (31,36). GST-tagged proteins bound to glutathione-agarose were equilibrated in cleavage/elution buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1 mM DTT, and 10% glycerol) and then incubated with tobacco etch virus protease overnight at 4°C to cleave protein from the GST tag. All proteins were purified by Mono Q anion exchange chromatography followed by S200 gel filtration chromatography in 20 mM Tris, pH 8.0, 150 mM NaCl, 10% glycerol, and 1 mM DTT. Eluted protein was concentrated to 20 -50 M working concentrations using a Millipore column concentrator, flash frozen in liquid nitrogen, and stored at Ϫ80°C.
Size Exclusion Chromatography-Analytical SEC was performed at 4°C on a Superdex 200 column in 20 mM Tris, pH 8.0, 150 mM NaCl ,and 1 mM DTT. Protein was injected at 25-30 M.
Limited Stokes Radius Measurements-The Stokes radius (R S ) was determined by analytical size exclusion chromatography using a Superdex 200 column equilibrated with 20 mM Tris, pH 8.0, 150 mM NaCl, and 1 mM DTT. Standard proteins were bovine carbonic anhydrase (R S ϭ 2.4 nm), bovine serum albumin (R S ϭ 3.5 nm), yeast alcohol dehydrogenase (R S ϭ 4.6 nm), sweet potato ␤-amylase (R S ϭ 5.4 nm), horse spleen apoferritin (R S ϭ 6.7 nm), and bovine thyroglobulin (R S ϭ 8.5 nm). The partition coefficient, K av , was calculated for all standards and ␣-catenin proteins used in this study. The Stokes radius was calculated from a standard curve of (ϪlogK av ) 1/2 versus R S .
Sucrose Density Gradient Centrifugation-Gradients of sucrose were made by layering sucrose dissolved in 20 mM Tris, pH 8.0, and 150 mM NaCl from 20 to 5% in 2.5% increments in 13 ϫ 63-mm ultracentrifuge tubes as described (44). Each layer was frozen in a dry ice/ethanol bath before the addition of the next layer. Tubes were stored at Ϫ80°C until use. Tubes were thawed overnight at 4°C to establish a gradient. 100 l of sample was layered on top and centrifuged in a Thermo Scientific Sorvall S100-AT rotor at 70,000 rpm (200,000 ϫ g) for 4 h at 4°C. All ␣-catenin proteins were loaded at concentrations Ն20 M. After centrifugation, 200-l fractions were collected and analyzed by SDS-PAGE. Gels were imaged on a LI-COR Biosciences scanner, and the percentage of protein in each fraction was measured in ImageJ. Plotted data were fit to a Gaussian curve to determine the peak fraction in Prism software. Standard proteins were bovine carbonic anhydrase (2.8S), bovine serum albumin (4.3S), yeast alcohol dehydrogenase (7.4S), sweet potato ␤-amylase (8.9S), and horse spleen apoferritin (16.6S). The sedimentation coefficient of ␣-catenin proteins was determined from a standard curve of sedimentation coefficient (S) versus fraction.
Molecular Mass Calculations-The molecular mass of ␣catenin proteins used in this study was calculated from the measured Stokes radius and sedimentation coefficient as described (34,45).
Actin Cosedimentation Assays-Chicken muscle G-actin (Cytoskeleton, Inc.) was incubated in 1ϫ actin polymerization buffer (20 mM HEPES, pH 7.5, 100 mM KCl, 2 mM MgCl 2 , 0.5 mM ATP, and 1 mM EGTA) for 1 h at room temperature to polymerize filaments. Gel-filtered ␣T-catenin or ␣T-catenin⅐ ␤-catenin heterocomplex was diluted to the indicated concentrations in 1ϫ reaction buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 0.5 mM ATP, 1 mM EGTA, 1 mM DTT, and 0.02% Thesit) with and without 2 M F-actin and incubated for 30 min at room temperature. Samples were centrifuged at 50,000 rpm (Ͼ100,000 ϫ g) for 20 min at 4°C in an S100-AT3 rotor. Pellets were resuspended in Laemmli sample buffer, separated by SDS-PAGE, and stained with Coomassie Blue. Gels were imaged on a LI-COR Biosciences scanner and measured and quantified in ImageJ. To determine the amount of bound protein, background sedimentation (no F-actin pellet) was first subtracted from cosedimentation (F-actin pellet). Bound protein across samples was then normalized to the F-actin pellet. The amount of bound protein was calculated from a standard ␣T-Catenin Is a Constitutive Actin-binding Protein curve created from the starting material. All binding data were processed with Prism software.
F-actin Bundling-Protein samples were prepared as for the actin cosedimentation assays and deposited on carbon grids. Samples were fixed in 2.5% glutaraldehyde, stained with 1% uranyl acetate for 1-3 min, and examined in a JEOL JEM-1011 transmission electron microscope. To quantify bundling, a 20 ϫ 20-m grid was overlaid on images, and the width of all bundles in four random squares on the grid was measured using ImageJ. The data were plotted and analyzed with Prism software.
Cross-linking Experiments-Purified ␣T-catenin ⌬ABD and ␣T-catenin V94D ⌬ABD were incubated with or without 1 mM BS3 (Thermo Scientific) in 20 mM HEPES, pH 7.4, 150 mM NaCl, and 1 mM DTT for 30 min at room temperature, separated by SDS-PAGE, stained with Coomassie dye, and imaged on a LI-COR Biosciences scanner.
GST Pulldown Experiments-Increasing amounts of ␣Tcatenin or ␣T-catenin V94D (1-15 g) were added to 15 g of GST-␤-catenin bound to glutathione-agarose in 20 mM Tris, pH 8, 150 mM NaCl, and 5 mM DTT. Samples were incubated with gentle mixing for Ͼ2 h at 4°C and then washed five times in PBS ϩ 0.05% Tween 20 and 5 mM DTT before elution in Laemmli sample buffer. Samples were separated by SDS-PAGE, stained with Coomassie dye, and imaged on a LI-COR Biosciences scanner.
Immunostaining and Confocal Microscopy-Cells were fixed in 4% paraformaldehyde in PHEM buffer (60 mM 1,4-piperazinediethanesulfonic acid, pH 7.0, 25 mM HEPES, pH 7.0, 10 mM EGTA, pH 8.0, 2 mM MgCl 2 , and 0.12 M sucrose), washed with PBS, blocked for 1 h at room temperature in PBS ϩ 10% BSA, washed three times in PBS, incubated with primary in PBS ϩ 1% BSA for 1 h at room temperature, washed three times in PBS, incubated with secondary in PBS ϩ 1% BSA for 1 h at room temperature, washed three times in PBS, and mounted in Fluoromount G (Electron Microscopy Sciences). F-actin was stained using Alexa Fluor-phalloidin (Invitrogen) and antibodies against ␣E-catenin (Enzo Life Sciences) or N-cadherin (Invitrogen). Cells were imaged on a Nikon Eclipse Ti inverted microscope outfitted with a Prairie swept field confocal scanner, Agilent monolithic laser launch, and Andor iXon3 camera using NIS-Elements imaging software. Maximum projections of 4-m image stacks were created for image analysis and presentation. For Pearson's r calculations, signal colocalization was measured between user-defined N-cadherin-positive AJ clusters and EGFP signals using ImageJ. Colocalization data were plotted and analyzed with Prism software.